To prospectively compare the accuracy of noncontrast magnetic resonance imaging (MRI) with indirect MR arthrography (I-MRa) of unexercised shoulders for diagnosis of superior glenoid labral lesions. Institutional Review Board approval and patient informed consent were obtained for this prospective study. Superior labral findings on shoulder MRI and unexercised I-MRa studies of 104 patients were correlated with findings at arthroscopic shoulder surgery. Two musculoskeletal radiologists independently reviewed the two sets of MR images while blinded to arthroscopic results. For each radiologist, the McNemar test was used to detect statistically significant differences between techniques. The superior labrum was intact in 24 and abnormal in 80 subjects. For detection of superior labral lesions by each radiologist, I-MRa was more sensitive (84-91%) than MRI (66-85%), with statistically significant improvement in sensitivity for one reader (p = 0.003). However, I-MRa was less specific (58-71%) than MRI (75-83%). Overall, accuracy was slightly improved on I-MRa (78-86%) compared with MRI (70-83%), but this difference was not statistically significant for either reader. Compared with noncontrast MRI, I-MRa was more sensitive for diagnosis of superior glenoid labral lesions. However, the diagnostic value of I-MRa in shoulders remaining at rest is potentially limited by decreased specificity of the technique.